374 MR T. C. BAILLIE ON THE 



(9) Difference in the emissive powers of the surfaces of the cooling bar and statical bar. 

 Some of the causes of such differences may be — 



(a) tarnish. 



(6) differences in the amount of polish. 



(c) difference in the surroundings or in the state of the atmosphere durinor 



the cooling and statical experiments. 



(d) differences in the radiation due to the temperature of the cooling bar 



always falling while that of the statical bar is steady. 



(10) Errors in the determination of the specific heat. 

 Of these the chief are — 



(a) the specimen used for this may not be a fair average specimen. 



(b) want of uniformity in its temperature when put into the calorimeter. 



(c) the calorimeter not receiving the whole of the heat supposed to be given 



out from the specimen. 



(d) changes in the thermal capacity of water with temperature as measured 



by the thermometer used (or, in the case of ice calorimeters, errors 

 in the value of the latent heat or other constants used). 

 Of these there is little difficulty in arranging the errors from (1), (2), (4), (6), (7), 



(8) to be small by using proper care and suitably arranging the apparatus or the bars 

 used. Serious error from (3) could be detected by taking a sufficient number of pro- 

 perly varied sets of readings. Small errors due to (9) (b) and (9) (c) are difficult to 

 avoid and it is impossible to discover their existence. (10) (c) is the most serious cause 

 of error in the ordinary method of mixtures. There is probably some error from this 

 cause even in Bunsen's calorimeter, as it usually gives lower values than other 

 methods. (10) (d) is unavoidable in all thermometric thermal measurements. (5) and 



(9) (d) are inherent to the method and are not to be avoided by the use of thermo- 

 electric junctions instead of thermometers. It is also impossible to estimate the errors 

 arising therefrom. In Angstrom's method errors from (10) affect the result in the same 

 way, and as all the temperatures measured are varying temperatures, errors of the same 

 sort as (5) and (9) (d) may occur. Angstrom's method is unreliable on other grounds. 

 It is essentially based on the assumption that the ratio of the conductivity to the 

 emissivity is constant. 



The values of the conductivity of copper found by Professor Tait were (reduced to 

 C.G.S. units) for good conducting copper 1*08 (1 + "0013£); for bad conducting copper I 

 71 (1 + *00j4£). The ratio of these values is independent of nearly every source of 

 error mentioned, and yet Dr Stewart (vide Trans. Roy. Soc, 1893, p. 569) found 1 "12 

 (1 — -00lt) ; while Kirchhoff and Hansemann (vide Wiedemann's Annalen, 9, p. 1 ; 

 13, p. 406) found '51 (1 +-0057t) both for pure copper. One has doubts about believ- 

 ing that the wide range of variations of these values is due only to differences in tin 

 specimens of metal used. I, therefore, determined to find the conductivity of the nickel 

 I had used by a method with fewer sources of error. 



